Public Health Ethics Advance Access published online on May 27, 2008
Public Health Ethics, doi:10.1093/phe/phn022
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
A Full-Pull Program for the Provision of Pharmaceuticals: Practical Issues
Australian National University
* Corresponding author: Centre for Applied Philosophy and Public Ethics (CAPPE), Menzies Centre for Health Policy, The Australian National University, LPO Box 8260, ANU Canberra ACT 2601, Australia. Tel.: +61 (0)2 6125 4355; Mobile: +61 (0)431 124 286; Fax: +61 (0)2 6125 6579; Email: michael.selgelid{at}anu.edu.au
 |
Abstract |
|---|
 Top Abstract IntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
Thomas Pogge has proposed a supplement to the standard patent regime whereby innovating companies would be rewarded in proportion to the extent to which their innovations lead to reduction of the global burden of disease. This paper argues that an expansion of this proposal—whereby provision of already existing medicines are incentivised in a similar way—would provide a more comprehensive solution to the healthcare situation in developing countries. It then considers the practical challenges that the implementation of such a proposal (expanded or otherwise) would entail. Though these include difficulties associated with disease burden metric, I argue that the most serious difficulties are associated with the problem of causal attribution. A basic idea underlying Pogge's proposal is that the disease burden reduction attributable to particular interventions can be determined. Theoretically speaking, in cases involving multiple interacting causal factors, there may be no fact of the matter regarding the extent to which disease burden reduction should be attributed to one intervention as opposed to another; and (even if workable practical solutions to this theoretical challenge can be met) the data requirements for implementation would be daunting.
|
Introduction |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
Thomas Pogge has proposed a supplement to the standard patent regime in the context of pharmaceuticals. Under the proposed new scheme, pharmaceutical companies would receive financial rewards as a function of the extent to which their innovations lead to reduction in the global burden of disease (Pogge, 2008).1 Pogge argues that an innovating company should, for any newly innovated product, be able to choose between two alternative reward schemes. Those who choose ‘track1’ would, as usual under the status quo, be able to set monopoly prices during the period of patent protection. Those who choose ‘track2’ would be rewarded in proportion to the health impact of the invention (e.g., over a period of 10 years), and the knowledge regarding the invention would be treated as a public good—i.e., made available for use by all.
Pogge's proposal partly aims to address what is commonly known as ‘the availability problem’—i.e., that only very few new pharmaceuticals for diseases primarily affecting the poor are brought into existence. Industry lacks sufficient financial incentive to engage in this kind of research under the status quo. If companies’ rewards are based on health impact, it would suddenly become profitable to develop new drugs for neglected diseases.
His proposal also aims to address what is commonly known as ‘the access problem’—i.e., that ‘available’ drugs are so often unaffordable (i.e., inaccessible) to the poor. If innovating companies are paid in proportion to the impact of their innovations in the way of disease burden reduction, then they will have motivation to make their drugs as affordable as possible, or perhaps even give them away for free. They will likewise have motivation to facilitate improvement of health care delivery/infrastructure in poor countries, to ensure that patients receive proper medication instructions, and so on (Pogge, 2008).
In what follows, I argue that an expansion of Pogge's proposal for patent reform—whereby health-promoting activities in general are incentivised in a similar way—would provide a more comprehensive solution to the healthcare situation in developing countries. I then proceed to examine the practical challenges that the implementation of such a proposal (expanded or otherwise) would entail. Though these include difficulties associated with disease burden metric, I argue that the most serious difficulties are associated with the problem of causal attribution. A basic idea underlying Pogge's proposal is that the disease burden reduction attributable to particular interventions can be determined. Theoretically speaking, in cases involving multiple interacting causal factors, there may be no fact of the matter regarding the extent to which disease burden reduction should be attributed to one intervention as opposed to another; and (even if workable practical solutions to this theoretical challenge can be met) the data requirements for implementation would be daunting.
|
A Full-Pull Program |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
The general idea of tying financial reward to impact for incentivising provision of innovative medicines for neglected diseases could, and arguably should, be further expanded to address the equally, if not more, important problem of poor peoples’ lack of access to already existing ‘essential medications’ which, though usually both inexpensive and off-patent, routinely remain unaffordable to those in need. The importance of increasing access to existing off-patent medications is well illustrated by infectious diseases such as tuberculosis (TB) and malaria, which (along with AIDS) are among the three biggest killers, annually claiming approximately 1.6 million and 1.2 million, respectively (Lopez et al., 2006: 8). Together these two diseases cause more deaths than AIDS (i.e., 2.1 million) each year. While antiretroviral treatment for AIDS in developing countries costs anywhere from $100 to $3002 per year or higher, a full course of TB medication, which would provide cure in the vast majority of cases, costs only $10–20. First-line TB medications have existed for over 40 years and are thus not eligible for patent protection. In 2004, new state-of-the-art malaria treatment, artemisinin combination therapy, cost only $2.40 wholesale and was expected to drop as low as $.50 or $1 within two years’ time (IOM, 2004). Though new, this treatment is not—and never was—subject to monopoly patent pricing.
The continuing large death toll from these diseases reveals that even low-cost non-patent-protected drugs are all-too-often out of reach for the poor in developing countries. In 1998, only 56 per cent of those in need worldwide had access to TB therapy recommended by WHO, and the rate was only 23 per cent just a few years earlier in 1995 (Lienhardt et al., 2003, 200). According to a recent report, even the previous standard (and still widely used) malaria medication, chloroquine, which costs only 10 cents per course of treatment, ‘is too [expensive] for the poorest of the poor in every endemic country’ (IOM, 2004: 61). One result of the unaffordability of chloroquine is that it has become almost useless in many regions due to drug resistance resulting (among other things) from patients’ inability to always complete a proper course of treatment. A real worry is that resistance to artemisinin will likewise emerge if access to this more expensive treatment is not improved. That resistance results from lack of complete access to TB treatment is another well-known, serious problem.
I have entered this brief discussion of malaria and TB, which are paradigm examples of diseases constituting the health problems of developing countries, to illustrate that patent reform may only provide a partial solution to the healthcare problems of such countries. The problem of patents is crucially important, because (1) patents lead to high prices that make medicines inaccessible to the poor, (2) standard patent incentives have proven insufficient to spur development of medical technologies most relevant to the needs of the poor, and (3) new drugs are needed, among other things, to deal with the growing problem of drug resistance. The importance of patents, however, is often overestimated. A recent study by Amir Attaran revealed that
in sixty-five low- and middle-income countries ... patenting is rare for [the] 319 products on the World Health Organization's Model List of Essential Medications. Only seventeen essential medications are patentable, although usually not actually patented, so that overall patent incidence is low (1.4 per cent) (Attaran, 2004: 155).
The fact that ‘over one third of the world's population lack access to essential drugs’ (WHO, 1999) (which are by definition ‘cost-effective’ and thus usually inexpensive), therefore, cannot be explained by patents alone. Contra Attaran, I do not deny that patents are important barriers of access to medication in poor countries (Selgelid and Sepers, 2006). The point is that lack of access to expensive patented medications is only part of the problem. The poor too often cannot afford cheap medicines either.
In the short-term, anyway, great strides would cost-effectively be made by merely increasing access to already existing, inexpensive, off-patent medications and other inexpensive items such as mosquito nets and sprays. Of the 18 million deaths in 2001 from ‘Group I’ causes—i.e., communicable diseases, maternal and perinatal conditions, and nutritional disorders—more than half were attributed to TB, malaria, diarrheal diseases, measles, lower respiratory infections, and malnutrition (Lopez et al. 2006: 8). In the vast majority of cases, these conditions can all be treated or prevented with already existing, inexpensive, off-patent medications and/or through improved sanitation, hygiene, nutrition, and education. UN Special Rapporteur Paul Hunt claims that ‘[i]mproving access to existing medicines could save 10 million lives each year’ (UN, 2007).
A virtue of Pogge's proposed solution to the problem of patents—i.e., for increasing both availability of and access to newly innovated drugs—is that it could be extended to address the lack of access to already existing medications. A straightforward extension of the initial scheme proposed by Pogge would simply be to reward companies that manufacture and distribute existing off-patent drugs as a function of the extent to which the drugs they manufacture and distribute lead to reduction in the global burden of disease (Selgelid and Sepers, 2006). The scheme could also be further extended beyond pharmaceuticals to cover things like mosquito nets, sprays, and diagnostic tests. Again we might imagine that the companies in question would even give such things away for free if the incentive scheme is sufficiently attractive (with regard to the amounts paid per unit of disease burden reduction).
The broadest extension of Pogge's idea would be to reward even those health-promoting interventions that do not involve medical technology at all. We could, that is, directly reward interventions involving improvement of things like education, water, and nutrition as a function of the extent to which such activities lead to reduction in the global burden of disease. If it is true that the greatest gains in global health would be achieved through interventions such as these, then we should not rely too heavily on the pharmaceutical industry for technological fixes to what are ultimately social problems. Given that global burden of disease studies show that ‘[u]ndernutrition is the single leading global cause of health loss’ (Ezzati et al., 2006: 247)—and that malnutrition is responsible for one-third of the disease burden in poor countries (Mason et al., 2003)—why not incentivise nutrition improvement by rewarding those who increase food provision to those in need as a function of the extent to which such activities lead to reduction in the burden of disease?
Because at least some extension to Pogge's initial proposal would be important if the aim is to more fully amend the healthcare situation in developing countries in a timely and cost-effective way, in what follows I will assume that a broader program aimed at incentivising at least both (1) innovation and access to new drugs and (2) access to already existing (off-patent) drugs is the goal when discussing how such a program might actually work in practice.
Before going on, however, I should point out two important ways in which the (extended) proposal would be distinct from another alternative to the status quo, which has recently gained much attention and popularity—that proposed by Michael Kremer and Rachel Glennerster (Kremer and Glennerster, 2004). Kremer and Glennerster are known for the idea of ‘pull-programs’—or Advance Purchase Commitments. The lack of new drug and vaccine development for neglected diseases, according to Kremer and Glennerster, can be explained by the way this kind of research is usually funded. The standard mechanism of ‘push funding’ involves donor or government provision of resources to stimulate research and development of medicines for neglected diseases. Though this may get relevant research going, innovating companies still lack incentive to complete large, expensive phase three clinical trials and bring new products to market given the lack of profits to be expected when those who need such products cannot afford to buy them. Pushed research on neglected diseases usually does not come to fruition. This problem can be addressed, according to Kremer and Glennerster, via Advanced Purchase Commitments: instead of providing push funding, wealthy-world governments and/or donor organizations should make legally binding up-front guarantees that they will buy sufficiently large numbers of desired new drugs or vaccines meeting specified criteria from companies that produce them. This virtual creation of a market would provide profit incentives that ‘pull’ innovation further down the research and development pipeline, and needed new medications for neglected diseases would finally be brought into existence.
While this is a good idea, Pogge's insight is that Kremer and Glennerster do not take the idea of pull incentives far enough. Even if Advance Purchase Commitments do bring new products to market, there may still be a lack of incentive for companies or others to deliver such products to those who actually need them. Kremer and Glennerster do not, that is, provide a sufficient solution to what is commonly known as ‘the last mile problem’.
By tying financial reward to impact, Pogge's proposal is that we should pull pharmaceutical products all the way to patients. Moreover, in comparison with Kremer and Glennerster's proposal, the extended proposal discussed above would pull more kinds of medications—i.e., already existing medications as well as those that are newly innovated—and perhaps nonmedical interventions as well. Pogge's scheme would pull more medications than the Kremer and Glennerster model even if the former is not extended beyond innovative products. Pogge's scheme would incentivise any innovation that leads to reduction in the global burden of disease (cost-effectively), while the Kremer and Glennerster model would only incentivise innovation of those medications specified by bureaucrats making Advance Purchase Commitment decisions. Because the extended scheme pulls more things—and pulls them further down the pipeline to patients—I refer to it as a ‘full-pull program for the provision of pharmaceuticals’, using the word ‘provision’ to indicate that both (1) innovation/availability problems and (2) access problems are explicitly addressed.
|
Political Feasibility |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
Though many will agree that the basic idea of a full-pull program has merits, some might object that it fails to provide a solution to the healthcare situation in developing countries insofar as it requires something already recognized as crucial to the solution of this problem—i.e., a huge influx of funding from wealthy-world governments and donor organizations (Selgelid and Sepers, 2006). If such large-scale funding dedicated to improving the health of the poor were available, the objection goes, then no revamping of the patent system would be necessary, because it would already be profitable for companies to invest in the kind of research and development that we are concerned about. If large sums of money were spent on improving global health, then the new market that the full-pull program aims to create would already exist—and the incentive to innovate would already be in place. Both the availability problem and the access problem might go away by themselves if wealthy countries were only more willing to pay for medicines needed by the poor: companies would invest in relevant R&D because there would be money in it, and the access problem would be solved because medicines and other health-promoting goods and services would be bought by the rich and provided to the poor. The primary problem with the status quo, the objection concludes, is merely that wealthy governments and donor organizations have proved insufficiently willing to spend on health improvement for developing countries.
It is here that the explicitly practical nature of Pogge's initial proposal becomes important, because Pogge believes that a full-pull program would generate its own political support. In particular, according to Pogge, we can realistically expect the pharmaceutical industry itself to lobby governments to put such a scheme into place and fund it. Though more funding from wealthy governments would be needed either way, we are more likely to get such funding if the pharmaceutical industry lobbies for it. We can expect this lobby because pharmaceutical companies would have much to gain and nothing to lose if a full-pull program is put into place, because—in the case of newly innovated drugs—they would be able to acquire monopoly pricing whenever they choose and because the addition of the alternative reward scheme would provide new ways of making money. In the case of already existing drugs, profit potential would likewise be expanded (i.e., especially for generic producers).
The pharmaceutical industry, however, could lobby governments to spend more on global health without lobbying for change in the current patent regime, and it would in that case too have much to gain and nothing to lose if such funding were in fact made available. With or without implementation of a full-pull program, increased wealthy-world spending on global health could provide new ways for pharmaceutical companies to make money. If there was only more money in it, developing drugs for neglected diseases could be profitable. One should thus ask why the pharmaceutical industry should be expected to support large-scale wealthy government funding of global health via implementation of a full-pull program rather than large-scale wealthy government funding of global health via the standard patent regime. Suppose that wealthy governments were to increase spending on developing country health improvement by X dollars. Wouldn't the pharmaceutical industry prefer that the X dollars be spent on monopoly priced drugs with high profit margins?
A reasonable response is that governments and the public would be more likely to support funding of global health via the full-pull program because it would provide more bang for the buck in the way of global health improvement given that (1) there would be no monopoly pricing for the drugs in question and that (2) spending would be tied to impact—and also because (3) a full-pull program would provide health benefits to citizens of wealthy countries themselves. If industry recognizes all of this, then it may support implementation of the full-pull program believing that wealthy governments would be more likely to fund it than they would be to fund improvement of global health under the current patent regime. An analogy may here be drawn to cases where a conservative party offers support to one of its more liberal over one of its more conservative candidates (despite the fact that the latter better represents its ultimate aims) when the former is more likely to gain wider popularity and thus win the election.
This argument regarding political feasibility provides an additional reason for extending the scheme to incentivise provision of already existing off-patent drugs as well as newly innovated ones. If there is a requirement that (in order to receive a reward based on impact) newly innovated drugs are used in cases where perfectly good already-existing ones would work, then the scheme would be less cost-effective (because newer drugs are more expensive when requisite research and development costs are taken into account). A more cost-effective program is likely to gain more political support.
|
Measurement Issues |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
The desirability and political feasibility of implementing a full-pull program also depends on whether or not, or how, such a program would actually work in practice, especially given the data requirements and measurement difficulties entailed. For such a program to work, we would need an adequate measure of the global burden of disease, sufficiently reliable predictions of what the global burden of particular diseases would be at particular times in the future in the absence of activities the scheme aims to incentivise and, finally, a way of determining the extent to which any reductions in the burden of particular diseases are the result of specific incentivised interventions.3 The last task is especially important because if companies are going to be paid as a function of the extent to which their drugs lead to reduction in the global burden of disease, then the impact of their drugs must be measured at the population level. This may be difficult, because the population impact of a new drug, or the population impact of making an already existing drug more widely available, is both theoretically and practically difficult to ascertain. The impact of greater drug provision will depend on other, natural and non-natural, factors affecting the population in question. Fluctuations in climate, nutritional status, water supply, education, economic status, behaviours related to health risk, availability of other drugs and medical care, and so on, may all affect the impact that increased provision of any particular drug will have in any given population.
The good news is that recent developments in the study of population health make implementation of a full-pull program more realistic than would have been the case at any previous time in history. The global burden of disease (GBD) studies purport to provide a comprehensive picture of the worldwide burden of over 100 diseases. The first study examined the burden of disease in 1990, and a more recent study examined the burden of disease in 2001 (Murray and Lopez, 1996; Lopez et al., 2006). It is safe to say that these have been by far the most comprehensive studies of their kind, and they have improved upon previous estimates of global population health by, among other things, providing internally consistent results. Because these studies enable—and in the first case explicitly provided—projections of the disease burden to be expected in the future, they have the potential to provide a basis for a full-pull program. The studies were in fact explicitly developed with the intention of facilitating distribution-of-resource policy making and assessment of the impact of healthcare interventions on populations.
Because the accuracy of such studies depends on the amount and quality of available data, on the other hand, improved data collection would be needed to increase confidence in both current estimates of the burden of disease and projected estimates of future disease burden. The best data comes from countries with the strongest vital registration systems—i.e., for the reporting and recording of each death and its cause, among other things. Unsurprisingly, however, such systems are usually weakest and often absent in developing world countries. There is less confidence in current disease burden estimates in poor countries, and the authors of the GBD studies urge ‘great caution’ in the use of their projections of future disease burden in places like sub-Saharan Africa in particular (Murray and Lopez, 1996: 331). For the purpose of a full-pull program, then, the data is weakest in the very places where it is wanted most. Projections are important because, for the plan to work, we would need to compare the actual burden of disease measured at times in the future with projections made ahead of time in order to base financial rewards on the extent to which the actual burden is less than would have been predicted (in the absence of incentivised innovations).
|
Disease Burden Metric |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
Data worries aside for the moment, another consideration regarding the suitability of the GBD studies as a basis for implementing a full-pull program is the question of disease burden metric, i.e., the units in which disease burden is measured. Given all the recent talk about ‘the global burden of disease’, it is important to note that, with regard to the GBD studies anyway, ‘disease burden’ is a technical term referring to something measured in DALYs, or ‘disability adjusted life years’ (Murray, 1994; Murray, 1996; Murray and Acharya, 1997). The concept of the DALY was in fact initially developed as part of the first GBD study. The DALY (species of QALY) is characterised as a ‘health gap measure’—it is meant to provide a common currency for measuring the amount of healthy life lost due to premature mortality and/or disability in comparison with some ideal. In the case of life expectancy the ideal (for women, anyway) is life expectancy in Japan, because this is where women live the longest, i.e., 83 years. If a woman succumbs to a disease that kills her 10 years earlier than a woman her age would be expected to die in Japan, then 10 DALYs are attributed to her premature death from this disease. If a person spends 10 years with a disability that reduces the functioning or quality of her life to half of perfect health, then five DALYs are attributed to the disability. If someone suffers 10 years with such a disability and then dies 10 years prematurely (by Japanese standards), then a total of 15 DALYs result—i.e., the five DALYs from disability and the 10 DALYs from premature mortality are simply summed.
Though some variant of the DALY may likely serve as an appropriate measure of the global disease burden (and/or its reduction) for a full-pull program, there are several controversial aspects of the GBD formulation of the DALY to be considered. An obvious question concerns the way in which the severity weights for disabilities should be determined. Different judgements about the severity of a condition like blindness are reached by those who are themselves blind, family members of the blind, the lay public, and health experts. While the GBD studies use the judgements of WHO experts to determine disability severity weights, which in turn are used to calculate disease burdens, one may reasonably question the objectivity of ‘expert judgments’ about such matters and thus the objectivity of the disease burden measures that result (Brock, 2004). I merely flag this issue and put it to the side for now.
Other controversial features of the DALY involve both age-weighting and time-discounting, whereby years of life lost are weighted unequally depending on one's age and the extent to which lost years of life would fall in the future. The 1990 GBD study, for example, weighted age according to a (somewhat) bell-shaped curve whereby a year of life during young adulthood counts for more than 1, peaking at 1.5 during one's early twenties, and a year of life counts for almost zero when one is close to birth or very old age. The 2001 GBD study, on the other hand, dropped the use of age weights in its DALY measurements. In both the 1990 and 2001 studies, a 3 per cent time-discounting rate was used, meaning decreasing importance gets attached to years of life lost as the years in question fall further into the future. Though much interesting debate surrounds the legitimacy of age-weighting and time-discounting, I will leave these issues to the side for now as their importance is only peripheral to the practical implementation of a full-pull program. I will say, however, that I personally agree with those (Anand and Hanson, 1997) who have argued that this kind of weighting should come into play, if at all, when making distribution of resource decisions rather than when making global disease burden measurements. If this is correct, then age-weighting and time-discounting should not be part of DALY calculations, but the architects and/or implementers of a full-pull program would still need to decide if greater financial awards should be attached to (averted) years of life lost at one age rather than another or to (averted) years of life lost sooner rather than later. The conclusion that these kinds of weightings should not be incorporated into DALY calculations does not imply that they have no role to play in resource allocation whatsoever.
|
Causation |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
The most important technical difficulties for the practical implementation of a full-pull program relate to problems of causation. Such problems occur at two separate levels. The first causation problem arises at the level of disease burden measurement. For the GBD studies, when a sick person dies her death is (in accordance with International Classification of Disease (ICD) guidelines) attributed to one particular disease and all years of lost life are attributed to that one disease—even in cases involving comorbidity. (In cases of comorbidity with HIV/AIDS and TB, for example, the ICD requires attribution of death to HIV/AIDS.) If a 40-year-old African woman dies of TB, then she dies roughly 43 years earlier than would be expected for a woman of her age in Japan—and thus 43 DALYs get attributed to TB given the way that disease burden gets measured. The reality of the situation, however, is that life expectancy is lower in Africa than in Japan. If the average life expectancy for 40-year-old women in Africa is 50 years and a woman in Africa dies of TB at 40, then in reality only 10 life years of her life are taken away by the TB infection that kills her. If this TB infection doesn't kill her, then the expectation is something else—or perhaps another TB infection—would kill her by age 50.4 The prematurity of her death by Japanese standards is thus, so-to-speak, overdetermined rather than being entirely the fault of this TB infection. This TB infection prevents the woman from making it to age 50, but a host of other factors (including the likelihood of other TB infections, if this one had not occurred) are responsible for the gap between 50 and 83.
While it may be appropriate to include the whole gap in time between her actual death and life expectancy in Japan to determine how many DALYs are lost (when this woman dies) for the purposes of GBD measurement, it would be wrong to attribute this entire 43-year gap to TB alone. If life-saving TB drugs are provided to a woman like this, then the global burden of TB would fall more than the overall GBD given the way that these burdens are measured.5 Reduction in the global burden of TB by amount X does not translate into reduction of the overall GBD by amount X as things are currently measured. We should thus not necessarily pay TB drug companies as a function of the extent to which their drugs lead to reduction in the global burden of TB if the mandate is to reward companies in proportion to their drugs’ reduction in the overall GBD. Determining the extent to which reduction in the global burden of TB leads to reduction in the total GBD, meanwhile, is difficult (for reasons related to those discussed in my second point below).
One way around this problem would be to change the way that the burden of a specific disease gets measured. If local life expectancy (rather than life expectancy in Japan) is taken into account, then in cases like that which we have been considering, 10 DALYs should be attributed to TB (when TB kills an African woman at age 40) and the remaining 33 years should be distributed among all the diseases (including TB) accounting for low life expectancy in Africa.6 Averting such a death by drug provision would then lead to a reduction of TB burden by 10 DALYs, and this would be equivalent to the reduction in GBD. Because an intervention's reduction in TB would thus be equivalent to its reduction in GBD, it would be simpler (assuming the requisite calculations can be made) to determine the rewards due to drug-providing companies—and the financial incentive to reduce the burden of particular diseases like TB would be clearer.
An alternative solution would be to change the way that GBD (as opposed to merely changing the way the burden of specific diseases) gets measured. Using local life expectancy (rather than life expectancy in Japan) as the target when determining how many years of life are lost when someone dies, we might say that only 10 years of life are lost when a 40-year-old African woman dies of TB. A reason not to calculate global disease burden itself using local life expectancy alone is that the death of a 40-year-old woman in Africa would then not appear to be as important as the death of a 40-year-old woman in Japan with regard to their impact on GBD—the former would count for only 10 DALYs while the latter would count for 43. Note, however, that (if financial rewards are based on the number of DALYs averted) on all three methods of measurement we have considered (i.e., the current method and the two alternative methods) one would have more incentive to save a 40-year-old in Japan from TB than one would have to save a 40-year-old in Africa from TB. Regardless of which of these three methods of measurement are used, saving the latter would have a greater impact on GBD. This does not mean, of course, that 40-year-old African women with TB would then be neglected. There are very few 40-year-old Japanese women with TB to save in the first place.
The second difficulty, which I consider to be the most serious obstacle to implementation of a full-pull program, is the problem of causal attribution. This is the problem of determining the extent to which any reduction in GBD—or the burden of any particular disease—is the result of one intervention as opposed to another. Let us suppose that a full-pull program is implemented and various health-promoting interventions take place as a result. New drugs and vaccines are developed and made available. Access to existing drugs is improved, and other health promoting activities involving education, water, and nutrition improvement take place. When the burden of a particular disease—say malaria—drops as a result, we would need to then determine the extent to which that drop is due to each particular drug or other intervention in order to determine the financial rewards owed to responsible parties. Making such a determination, however, involves serious theoretical and practical difficulties.
Theoretically speaking, one might doubt that there is any fact of the matter regarding the extent to which a drop in disease burden breaks down into component causes in cases where multiple interacting causal factors are involved. For simplification, we can illustrate this problem at the level of the individual. Imagine that someone who would have died from malaria ends up living because she receives a partially effective vaccine and gains access to a mosquito net. Even with perfect data availability, it may then be dubious to say that it was either the vaccine or the mosquito net that saves her life—and it may be dubious to say that there are some numbers X and Y such that her survival is X% caused by the vaccine and Y% (where Y = 100 – X) caused by the net. To illustrate this point an analogy can be made with genetics. If a person with a gene that increases her chance of dying of cancer actually gets and dies of cancer, it would be wrong to think that there are some numbers X and Y such that her death is X% caused by the gene and Y% caused by environment. It is the interaction between one's genes and one's environment that leads to her cancer/death, and there is little to be said about the quantitative extent to which her death is due to one cause as opposed to the other (Kaplan, 2000). (Heritability values tell us something very different: i.e., the percentage of phenotypic variation in a population that is due to genetic variation, as opposed to environmental variation, in the population in question.) I take this to be a well-established important message in the literature on genetic determinism, and it appears that this point about causation of disease by genes also applies to causation of disease avoidance by things like drugs, vaccines, and mosquito nets.
At the population level, in any case, this type of theoretical problem is dealt with in the science of public health by means of counterfactual analysis (Murray and Lopez, 1999).7 To attribute the drop in a particular disease to one particular health intervention, say the provision of a particular new drug to the population, we ask how many more deaths there would have been if, holding other things equal, the drug had not been provided8—and we then take this to be the number of averted deaths attributable to the drug in question. This is not a perfect solution to the problem, however, because counterfactual analysis may not provide what is known as ‘additive decomposition’—i.e., the number of lives attributed to different causes need not add up to the total number of lives that are saved when causal attribution is determined this way.
Case 1: Post-Intervention Baseline
To illustrate, imagine two drugs: Drug A saves 20 per cent of AIDS sufferers who receive it when it is given in isolation and Drug B saves 30 per cent of AIDS sufferers who receive it when it is given in isolation. Furthermore, imagine that there is a synergistic effect between the two drugs such that 100 per cent of those who receive both drugs are saved. Now imagine a population of 100 AIDS sufferers who receive both drugs and are thus all saved as a result. The burden of AIDS thus drops by 100 deaths.9 How many of these averted deaths should be attributed to Drug A and Drug B, respectively? Using counterfactual analysis, we determine how many averted deaths should be attributed to Drug A by asking how many people would have died if it had not been provided, other things being equal. If it had not been provided, then the 100 people would have received Drug B alone. 30 of them would have lived, and there would be 70 more deaths than the scenario in which both drugs are given. We thus attribute 70 averted deaths to Drug A. Using counterfactual analysis, we determine how many averted deaths should be attributed to Drug B by asking how many people would have died if it had not been provided, other things being equal. If it had not been provided, then the 100 people would have received Drug A alone. Twenty of them would have lived, and there would be 80 more deaths than the scenario in which both drugs are given. We thus attribute 80 averted deaths to Drug B. Note, however, that the sum of the number of averted deaths attributed to Drug A and the number of averted deaths attributed to Drug B—i.e., 70 + 80 = 150—exceeds the total number of deaths actually averted, i.e., 100. A workable way of divvying up the financial reward due to the 100 deaths actually averted is nonetheless available. We can simply divide the total reward for the 100 deaths averted using the ratio of 70 to 80—i.e., 7 to 8—for Drug A and Drug B, respectively. If the reward scheme was such that $1 is paid per death averted, then the total payment should be $100. Using the 7-to-8 ratio, $47 would thus be paid to the providers of Drug A and $53 would be paid to the providers of Drug B.
Case 1: Pre-Intervention Baseline
It should be noted, however, that the numbers that result depend on the ‘actual’ baseline from which the counterfactual analysis is conducted. Above I assumed the ‘actual’ baseline was the post-intervention actual situation where both drugs were provided. We might alternatively use the pre-intervention situation, before either drug was provided, as the ‘actual’ baseline. In this case the baseline is one where 100 people die of AIDS. We would then determine the number of averted deaths that should be attributed to Drug A by asking how many lives would be saved in the counterfactual situation where Drug A alone is provided to the whole population. If the 100 peo- ple are provided with Drug A alone, then 20 people would be saved; and we attribute 20 averted deaths to Drug A. We would likewise determine the number of averted deaths that should be attributed to Drug B by asking how many lives would be saved in the counterfactual situation where Drug B alone is provided to the whole population. If 100 people are provided with Drug B alone, then 30 people would be saved; and we attribute 30 averted deaths to Drug B. In this case the sum of averted deaths attributed to Drug A and the number of averted deaths attributed to Drug B—i.e., 20 + 30 = 50—is less than the total number of deaths actually averted when the whole population receives both drugs, i.e., 100. We could nonetheless in this case use the ratio of 20 to 30—i.e., 2 to 3—to determine how the $100 reward should be divvied up when both drugs are provided and 100 lives are saved. In this case, using a 2-to-3 ratio, $40 would go to the providers of Drug A and $60 would go to the providers of Drug B. An advantage here is that the numbers match the comparative efficacies of the two drugs when they are given in isolation.
Case 2: Time-of-Intervention Baseline
I have run through implications of using the pre-intervention scenario as the baseline in order to show that the solution generated above (using the post-intervention scenario as the baseline) did not capture any fact of the matter regarding the way in which rewards should be divided. This point is strengthened when we take into account that there are additional baselines that might be chosen. We could, for example, use the actual situation at the time each given intervention takes place to determine the number of averted deaths that should be attributed to that intervention. Imagine, for example, that Drug A is provided to the 100 AIDS sufferers at time t1 and that Drug B is provided later at time t2. Let us further assume, to keep things simple for the purpose of illustrating the problem, that no one dies between the time that Drug A is provided and the time that Drug B is provided. Using the actual situation at the time each drug is provided as the actual baseline, the baseline for Drug A is the situation where 100 people die of AIDS. If Drug A is counterfactually provided, then 20 will live and 80 will die. 20 averted deaths are thus attributed to Drug A. The baseline for Drug B is the status quo at the time it is provided—i.e., Drug A but not Drug B is being provided so that 20 would live and 80 would die. If Drug B is counterfactually provided, then 100 will live and no one will die. There would be 80 fewer deaths and so 80 averted deaths are attributed to Drug B. Using the ratio of 20 to 80—or 2 to 8—$20 would be awarded to the providers of Drug A and $80 would be provided to the providers of Drug B. An advantage of using the actual situation at time of intervention as the baseline is that we in this case achieve additive decomposition. The averted deaths attributed to each intervention sum to the total number of averted deaths: 20 + 80 = 100. A second advantage is that parties would be rewarded for the actual value of what they do at the time that they actually do it (given other things that are already taking place).
Case 3: Time-of-Intervention Baseline
A disadvantage of using such a baseline, however, is that it may sometimes lead to counterintuitive results and/or create perverse incentives—i.e., to delay or withdraw intervention. Imagine, for example, that Drug A saves 40 per cent if given in isolation and that Drug B saves 5 per cent if given in isolation and that 100 per cent are saved if both drugs are given in combination. Imagine that Drug A is given at time t1 and that Drug B is given later at time t2. Using the actual situation at the time of intervention as the baseline, 40 averted deaths would be attributed to Drug A and 60 to Drug B. Such a result would be counterintuitive because Drug B is hardly efficacious on its own in comparison with Drug A. If Drug B is given first and Drug A is given later, then five averted deaths would be attributed to Drug B and 95 to Drug A. The general problem is that the benefits of synergistic effects would all be attributed to the later providers when the time-of-intervention baseline is used. Given the implications of this with regard to financial rewards, the providers of Drug A would have (perverse) incentive to hold off on initial provision of Drug A (if they are aware that Drug B is in the pipeline) or to actually stop provision of Drug A and then provide it again at time t3 when the status quo is that the population is receiving Drug B alone. (On this latter course of events, we would no longer have additive decomposition.)
Case 4: Post-Intervention Baseline
Counterintuitive results likewise result from the first two baselines considered above. Let's again consider the post-intervention baseline. Imagine that Drug A saves 5 per cent when given in isolation, that Drug B saves 5 per cent when given in isolation, that drugs A and B save 100 per cent when given in combination, and that Drug C saves 100 per cent when given by itself. Imagine a population of 200 that will die of AIDS if no intervention takes place and that 100 people receive drugs A and B in combination and that 100 (others) receive Drug C. 200 deaths are averted. If Drug A is not given, then 100 people would receive Drug B alone and five of them would be saved and 100 people would receive Drug C alone and all of them would be saved. Ninety-five more people would die, and so 95 averted deaths would be attributed to Drug A. On the same analysis, 95 averted deaths would be attributed to Drug B. If Drug C were not given, then 100 people would receive drugs A and B in combination, and they would all live. 100 people would die. One hundred deaths would thus be attributed to Drug C. The ratio for financial reward divvying would then be 95 to 95 to 100 for drugs A, B, and C, respectively. This is counterintuitive because the providers of drugs A and B would be rewarded almost as much as the providers of Drug C despite the fact that drugs A and B do so little when given in isolation in comparison with Drug C. The general problem here is that the full benefits of synergistic effect are attributed both to Drug A and to Drug B—rather than being distributed between the two. Ignoring the fact that drugs A and B have limited efficacy when given in isolation, one would intuitively think that the ratio should be 50 to 50 to 100 in a case like this. Given that the A/B combination saves the same number of lives as Drug C, those providing the A/B combination should between them receive the same reward as the providers of Drug C.
Case 4: Pre-Intervention Baseline
For a counterintuitive result using the pre-intervention baseline, consider the following example. Imagine, again, that Drug A saves 5 per cent when given in isolation, that Drug B saves 5 per cent when given in isolation, that drugs A and B save 100 per cent when given in combination, and that Drug C saves 100 per cent when given by itself. Imagine a population of 200 that will all die of AIDS if no intervention takes place and that 100 people receive drugs A and B in combination and that 100 (others) receive Drug C. 200 deaths are averted. Using the pre-intervention scenario as the baseline, we determine how many averted deaths to attribute to Drug A by asking how many people would be saved if Drug A is given alone. In that case, five would live and so five averted deaths are attributed to Drug A. By the same analysis, five averted deaths would be attributed to Drug B. If Drug C were given alone, then 100 would live and so 100 averted deaths are attributed to Drug C. In this case the ratio would be 5 to 5 to 100 to drugs A, B, and C, respectively. The problem is then that relatively little incentive would motivate provision of the A/B combination. Intuitively one would have thought that the amount owed to the providers of drugs A and B would sum to the amount owed to the providers of Drug C. The general problem is that the use of a pre-intervention baseline does not in a case like this account for synergistic benefits—and so incentives are lower than they should be.
We have found that, for each of the baseline scenarios described, cases can be constructed that yield counterintuitive results and that it is possible to construct cases where substantially different results obtain from use of different baselines. Be that as it may, a (perhaps arbitrary) decision to use one baseline or another would hopefully be antecedently agreeable to stakeholders and sufficient to provide a clear enough incentive scheme to motivate the kinds of interventions we want to take place if a full-pull program is implemented. The time-of-intervention baseline is perhaps preferable simply because the incentive it provides for any given intervention is directly proportional to the value of that intervention (in terms of disease burden reduction) at the time the intervention takes place. In the case of the time-of-intervention baseline, furthermore, negotiation between drug providing companies themselves would likely resolve the possibly counterintuitive results and perverse incentives that have been mentioned. In Case 3, for example, rather than actually delaying or withdrawing its intervention, the providers of Drug A could strike a bargain with the providers of Drug B regarding the way that the total reward for total disease burden reduction will be distributed when both drugs are given in combination. Providers of Drug A would have real bargaining power in a scenario like this, because the reward due to providers of Drug B would be less if Drug A is delayed or withdrawn. Rules could, alternatively, be put into place to prevent providers of Drug A from being rewarded for their second round of intervention in a population if they pull out and then reintervene once Drug B is being provided to that population in scenarios like Case 3. To prevent delay of intervention, special bonuses could be offered to early interveners. Or rules might (perhaps arbitrarily) specify that rewards due to synergistic benefits are divided equally between providers of drugs that produce synergistic effects—or that rewards due to synergistic benefits are divided among providers according to the ratio of their drugs’ efficacies when used in isolation. There may thus be various practical solutions to the theoretical problem of causal attribution.
We should not, however, underestimate the difficulties highlighted by the case analyses above. Though the considered cases all raise questions about how to attribute disease burden reduction to different drugs, the same kind of difficulty will arise when other disease reduction interventions (such as water or nutrition improvement) take place in a population. Because there would often be synergistic effects between pharmaceutical and non-pharmaceutical interventions, the puzzles associated with synergistic effects would by no means be limited to exceptional cases involving drug cocktails.
|
Conclusion |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
A final problem is that conducting the kind of counterfactual analysis described would in practice require an enormous amount of data. Unless there exist populations that are in all relevant respects exactly the same as those we are interested in—except for the fact that the former lack the particular intervention for which we want to determine the disease-reduction effects—the counterfactual scenarios would have to be modelled in order to ascertain the needed numbers. To make a long story short, the data requirements of such modelling would be daunting to say the least. Implementation of a full-pull program would pose tremendous data challenges for (better) measuring the actual global burden disease and making projections about the future GBD (as described above). But the data requirements for counterfactual solutions to causal attribution problems may be even greater.
Data-limitation may not itself, however, provide reasons to reject the full-pull program proposal. Given the importance of the problem that such a scheme aims to address, a coalition of willing countries and/or funding bodies could decide to proceed with the best scientific evidence available, despite the inevitable uncertainties involved. The problem is extremely important, so we might simply do the best we can.
Measurement worries can, in any case, be further addressed by improving data availability in poor countries; and there are good reasons for improving data collection in poor countries whether or not a full-pull program is ever adopted. It is already widely acknowledged that better data is needed for disease impact assessment, priority setting, and health policy making in general. Given global concerns about emerging infectious diseases—and the ease with which they can spread internationally in a globalised world, as illustrated by SARS and avian influenza—wealthy nations themselves have independent interests in improving health surveillance in poor countries—and the new WHO International Health Regulations require that improved surveillance be put into place. Economists, furthermore, hold that health surveillance should be treated as a global public good (Zacher, 1999). We might realistically hope, in any case, that poor countries would themselves be willing to work to improve data availability in their countries if this could enable the major health improvements promised by a full-pull program. Improvement of vital registration systems and other means of data collection could, finally, be mandated by a treaty or compact between participating countries.
So long as companies were sufficiently confident that their successful efforts would be rewarded despite imperfections in measurement, they would have motivation to increase provision of medicine. So long as there is sufficient confidence that real progress in disease burden can be made at a reasonable cost despite imperfections in measurement, governments and their citizens could feel justified in funding the project. Whether or not a full-pull program would provide a politically feasible, ‘workable’ solution to the healthcare situation in developing countries depends on the extent to which adequate measurements can be made with existing or realistically attainable data. This remains to be seen.
|
Acknowledgements |
|---|
An earlier version of this paper was presented at the 8th World Congress of Bioethics in Beijing, China, in August 2008. I have especially benefited from discussion with—and/or comments on previous versions of this paper from—Bengt Brülde, Richard Cash, Kieran Donaghue, Mira Johri, David Mollica, Thomas Pogge, Michael Ravvin, Daniel Star, Christopher Wellman, and the editors of Public Health Ethics.
|
Notes |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
1 A similar proposal has been advanced by Aidan Hollis (2008).
2 All dollar figures in this paper refer to US currency.
3 Pogge suggests that we may not need to disentangle reductions in disease burden arising from "natural causes"—such as the weather—from those arising from incentivised activities. We could, he suggests, simply allow companies to benefit or suffer from such fluctuations as luck would have it. But we surely must disentangle disease burden reductions arising from various human activities. If local government provision of clean water and the provision of a new drug both lead to reduction in a particular disease in a particular population, we should not reward the pharmaceutical company for disease reduction resulting from the government's (independent) action.
4 For useful discussion of these and related issues, see Williams (1999); Murray and Lopez (2000); and Williams (2000).
5 Forty-three years of lost life would be attributed to TB if the woman died at age 40. If we imagine that she is saved by TB drugs but then still dies at the age of 50 from something else, then the burden of TB would drop by 43 years but the GBD would drop by only 10 years.
6 The question about how exactly the years between local life expectancy and 83 should be distributed among the diseases accounting for low life expectancy is beyond the scope of this paper. For the purpose of simplicity, I am assuming that life expectancy in Africa would remain the same if TB were eliminated. A more complicated solution (if this assumption is incorrect) would be to attribute the gap between the age at which a person dies (from TB) and what life expectancy would be in Africa if TB were eliminated to TB and then attribute (distributively) the gap between that age and 83 to the remaining causes of low local life expectancy.
7 Counterfactual analysis involves comparison of an actual scenario with a counterfactual one.
8 This kind of question is answered by looking at a population that is similar—except for the fact that the intervention in question has not taken place—and using mathematical modelling to account for any (other) relevant differences.
9 For the sake of simplicity, this discussion will speak of disease burden in terms of numbers of deaths rather than DALYs.
|
References |
|---|
|
TopAbstractIntroductionA Full-Pull ProgramPolitical FeasibilityMeasurement IssuesDisease Burden MetricCausationConclusionNotesReferences |
|---|
-
Anand S., Hanson K. Disability-Adjusted Life Years: A Critical Review. Journal of Health Economics (1997) 16:685–702.[CrossRef][ISI][Medline]
Attaran A. How Do Patents and Economic Policies Affect Access to Essential Medicines in Developing Countries? Health Affairs (2004) 23:155–166.
Brock D. W. Ethical Issues in the Use of Cost Effectiveness Analysis for the Prioritisation of Health Care Resources. In: Public Health, Ethics, and Equity—Anand S., Peter F., Sen A., eds. (2004) New York: Oxford University Press. 201–223.
Ezzati M., Hoorn S. V., Lopez A. D., Dannaei G., Rodgers A., Mathers C. D., Murray C. J. L. Comparative Quantification of Mortality and Burden of Disease Attributable to Selected Risk Factors. In: Global Burden of Disease and Risk Factors—Lopez A. D., Mathers C. D., Ezzati M., Jamison D. T., Murray C. J. L., eds. (2006) Washington, DC: The World Bank and Oxford University Press.
Hollis A. ‘A Comprehensive Advance Market Commitment: A Useful Supplement to the Patent System. Public Health Ethics (2008) 2:in press.
Institute of Medicine (IOM). Saving Lives, Buying Time: Economics of Malaria Drugs in an Age of Resistance (2004) Washington, DC: National Academies Press.
Kaplan J. M. The Limits and Lies of Human Genetic Research (2000) Chap. 3. New York: Routledge.
Kremer M., Glennerster R. Strong Medicine (2004) Princeton, NJ: Princeton University Press.
Lienhardt C., Ogden J., Sow O. Rethinking the Social Context of Illness: Interdisciplinary Approaches to Tuberculosis Control. In: The Return of the White Plague: Global Poverty and the ‘New’ Tuberculosis—Gandy M., Zumla A., eds. (2003) London: Verso. 195–206.
Lopez A. D., Mathers C. D., Ezzati M., Jamison D. T., Murray C. J. L. Global Burden of Disease and Risk Factors (2006) Washington, DC: The World Bank and Oxford University Press.
Mason J. B., Musgrove P., Habicht J. P. At Least One-Third of Poor Countries’ Disease Burden is Due to Malnutrition. (2003) Working Paper No. 1, Disease Control Priorities Project. Bethesda, Maryland: Fogarty International Center, National Institutes of Health.
Murray C. J. L. Quantifying the Burden of Disease: The Technical Basis for Disability-Adjusted Life Years. Bulletin of the World Health Organization (1994) 72:429–445.[ISI][Medline]
Murray C. J. L. Rethinking DALYs. In: The Global Burden of Disease: A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries and Risk Factors in 1990 and Projected to 2020—Murray C. J. L., Lopez A. D., eds. (1996) Cambridge, MA: Harvard School of Public Health. 1–98. Published by Harvard School of Public Health on behalf of the World Health Organization and The World Bank.
Murray C. J. L., Acharya A. K. Understanding DALYs. Journal of Health Economics (1997) 16:703–730.[CrossRef][ISI][Medline]
Murray C. J. L., Lopez A. D. The Global Burden of Disease: A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries and Risk Factors in 1990 and Projected to 2020 (1996) Cambridge, MA: Harvard School of Public Health. Published by Harvard School of Public Health on behalf of the World Health Organization and The World Bank.
Murray C. J. L., Lopez A. D. On the Comparable Quantification of Health Risks: Lessons from the Global Burden of Disease Study. Epidemiology (1999) 10:594–605.[CrossRef][ISI][Medline]
Murray C. J. L., Lopez A. D. Progress and Direction in Refining the Global Burden of Disease Approach: A Response to Williams. Health Economics (2000) 9:69–82.[CrossRef][ISI][Medline]
Pogge T. World Poverty and Human Rights: Cosmopolitan Responsibilities and Reforms (2008) 2nd, expanded edn. Chap. 9. Cambridge, MA: Polity Press.
Selgelid M. J., Sepers E. Patents, Profits, and the Price of Pills: Implications for Access and Availability. In: The Power of Pills: Social, Ethical and Legal Issues in Drug Development, Marketing and Pricing Policies—Cohen J. C., Illingworth P, Schüklenk U., eds. (2006) London: Pluto Press. 153–163.
United Nations (UN). UN Independent Expert Launches Draft Human Rights Guidelines for Pharmaceutical Companies. (2007) Press Release, 19 September. Available at: http://www.unhchr.ch/huricane/huricane.nsf/view01/497E81A16B31B9C8C125735B-0059D7B0?opendocument [Accessed 23 December 2007].
Williams A. Calculating the Global Burden of Disease: Time for a Strategic Reappraisal. Health Economics (1999) 8:1–8.[CrossRef][ISI][Medline]
Williams A. Comments on the Response by Murray and Lopez. Health Economics (2000) 9:83–86.[CrossRef][ISI][Medline]
WHO. Report on Infectious Diseases: Removing Obstacles to Healthy Development (1999) Geneva: WHO Publications. Available at http://www.who.int/infectious-disease-report/pages/textonly.html [Accessed 29 December 2007].
Zacher M. W. Global Epidemiological Surveillance: International Cooperation to Monitor Infectious Diseases. In: Global Public Goods: International Cooperation in the 21st Century—Kaul I., Grunberg I., Stern M. A., eds. (1999) New York: Oxford University Press. 266–283. Published by Oxford University Press on behalf of The United Nations Development Program.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. Hollis The Health Impact Fund: A Useful Supplement to the Patent System? Public Health Ethics, July 1, 2008; 1(2): 124 - 133. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||